Antrlyst, October 1996. Vol. 121 (1367-1372) 1367 Cross-sections of Spectrochromatograms for the Resolution of Folpet, Procymidone and Triazophos Pesticides in High-performance Liquid Chromatography With Diode-array Detect ion J. L. Martinez Vidal, P. Parrilla, M. Martinez Galera and A. Garrido Frenich Depar-tnici?t of' Arwlytical Chemisti-y, 7 J ~ i i ~ r s i t y of' Alnr~iY'u, 04120 AImci-fa, Spa in The rapid-scanning photodiode array detector generates a considerable amount of data in HPLC, such as the three-dimensional (A, h, t ) matrix, but requires improvements in data analysis methodology to utilize all the available information. In this paper, a graphical technique is used for improving the selectivity of HPLC analysis, using the available spectrochromatographic information in both the time and wavelength domains.The technique consists in performing cross-sections through the data matrix to obtain selective analytical information for each of the analytes. In order to demonstrate the validity and simplicity of the method it has been applied in the simultaneous determination in mixtures of the three pesticides folpet, procymidone and triazophos. The procedure was applied with satisfactory results in the determination of these pesticides in groundwater at ppb levels after solid-phase extraction with C18 cartridges. Keywords: Pestic.idcs; f d p e t , procymidoiie nix/ ti.iiizophos; M'atei-; h ig Ii -peifoi-nmiic*e 1 iqii id chi-onza tog rap h y ; diode - ai*ra~ detection; c,r.o.ss-set.tions Introduction The monitoring of pesticides i n different environmental niatri- ces is an analytical problem of growing importance.Ideally, the deployment of few. inexpensive multi-residue (MR) methods would facilitate the rapid identi f'ication and quantification of a wide range of pesticides at the required sensitivity limit, in response to legislation in many countries. Different techniques have been applied in the determination of pesticides, mainly employing CCI-I and HPLC4-7 with a variety of detectors. However, the properties of relatively new classes of pesticides, such as phenylureas, phenoxy acids, carbainates and quaternary amines, make these more suitable to HPLC than to GC. There are several detection methods for HPLC analysis, such as UV/VIS, refractive index, electrochemical, fluorescence and chemiluminescence.UV/VIS-based diode-array detection methods (DAD) is one of the most commonly used multi-wave length detection methods in HPLC to gain more analytical inform at i on about t lie in i x t ures of i n t crcs t .8- I 0 The use of HPLC-DAD has several advantages beyond simple component identification. In the temporal domain, the analysis time can be shortened because the wavelength dimension allows the analyst to observe all UV/VTS-absorbing components during a single elution. In the spectral domain, an improvement in the detection level is gained owing to the availability of the total UV/VIS spectrum. In multicomponent mixtures, where the analytes are not resolved by the column but where spectral overlap is minimal, the analytes can be determined simultaneously by monitoring each component at a wavelength that is free of interference. Hence the analyst can obtain rnultiwavelength chromatograms from a single analysis of one sample in order to resolve overlapping signals.On the other hand, in the case of complex mixtures, where all the analytes comprising a single elution profile and where spectral overlap is severe," the technique of obtaining chromatograms at different wavelengths is not adequate t o resolve overlapping peaks. Typical examples of overlapped peaks can occur if new pesticides have to be checked with an established MR method or if interferents from complex sample matrices are co-eluted. In this situation, where the overlapping signals do not permit the analysis of all analytes in a single chromatographic run, it is possible to modify the MR method or to apply chemometric techniques in order to extract useful information from the overlapped region.The first solution is not the most adequate because of the great cost involved in developing a new method. Therefore, the second solution is usually chosen, but a single compromise detector wavelength has to be selected to apply the majority of chemometric methods. Moreover, one intrinsic advantage of multiwavelength detection in HPLC is that data are directly available in digitized form for storage and software manipulation, by a number of experimental procedures for the characterization of unresolved peaks. Also, many of the algorithms developed for analytical spectroscopy can be used for data analysis in HPLC-UV/VIS.l4-I9 The objective of this work was to apply a methodological approach to extract selective analytical information from the data generated by HPLC-DAD. The method consists in the generation of cross-sections through the three-dimensional ( A , h, t ) matrix providing the optimum signal relative to possible interfering analytes to gain selectivity. The proposed method allows the combined use of spectral and chromatographic information for the deconvolution of overlapping peaks, opening up new prospects for DAD in HPLC. This method has been applied in the determination of the pesticides folpet, procymidone and triazophos simultaneously present in synthetic mixtures. Generally, these compounds are determined by chromatographic methods, either GC or HPLC.Methods for the GC determination of folpet, procymidone and triazophos have been reported with electron-capture,2" ther- nioionic N-P," tlame ionization2* and mass detection.2? HPLC methods have been used with UV24325 and mass detection.26 The procedure was applied to the determination of these pesticides in groundwater at ppb levels after solid-phase extraction (SPE) with CIS cartridges.I368 Aiiirlyst, Octohci. 1996, Vol. 12 I Experimental R eage it ts HPLC-grade solvents were used. The pesticide standards (pestanal quality), summari& i n Table I , were obtained from Riedel-de HaEn (Seelze, Gerrnany). Solid \tandads were dissolved in acetonitrile (ACN) and \tored at 4 "C in the dark. where they were stable for \everal months. Working solutions were prepared daily by appropriate dilution with ACN.Mobile phases were de-gassed with heliuni prior and during u\e. Distilled water was obtained I.rom a Millipore (Bedford, MA, USA) Milli-Q water purification system. All solvents and samples were filiered through Millipore rneinbrane filters before iiijection into the column. Prepacked Sep-Pak C I x cartridges containing 360 mg of' C I chemically bonded silica (Waters, Milford, MA, USA) were used. Apparutus A Watery Model 990 liquid chromatographic system was used, equipped with ;I Model 600E constant-flow pump. a Rheodyne six-port injection valve with a 20 pl sample loop and a Model 990 photodiode-array detector. The spectral resolution used was 1.4 nm per diode in the range 200-280 nm. HPLC separations were carried out using ;t Eiypersil Shandon Green Env. I SO X 3 nirn id ( 5 pin particle sirre) C l x column.Sojtware A coinpati ble personal computer provided with a 486 micro- processor and mathematical coprocessor was used for acquisi- tion ancl treatment of the data. 'The liquid chromatographic system allow\ the acqui5ition of' ;t series of chromatograms at different wavelengths. The Waters 99 I software controlling the instrument generates a three-dimensional file ( A , A. f) in binary format. Then, the three-dimen~ional tile is converted into a series of / I individual spectra. each corresponding to an absorption spectrum, acquired tit a different time, with the ASCII converter included in the Waters 99 1 program. The resolution used i n the time doinain is 1.4 c .A converter program i n BASIC was used to transform the two-dimensional files into ASCII format tor the software packages SURFER and GRAPHER.27 The three-dimensional spectrochromatc,grams are obtained and presented as isometric plots (A, h, t). Alternatively. the data are presented as a contour plot in both the time and wavelength dimensions, by linking points of equal intensity to form the contour map. The SURFER program permits the generation of cross-sections and shows the trajectory followed in the contour or isometric plot. Using GRAPHER software, cross-section data are plotted to produce a profile from the two-dimensional data projection [A-f(h, t ) ] . When the data are plotted, the absorbance value is plotted a s the y coordinate. HPLC Operating Conditions The following conditions were used: flow rate, I ml min-1; chart speed, 0.5 cm min-1; detector sensitivity, 0.02 a.u.f.s.; and column at room temperature.The solvent programme was as follow\: initially 2 min isocratic with water-ACN-MeOH (56 + 27 + 17). followed by a 20 min linear gradient water-ACN- MeOH ( 5 + 90 + 5 ) ; an additional period of 1 0 min of gradient programme was sufficient to return the \ystem to the initial conditions for subsequent analysis runs. The solvents were filtered daily through a 0.45 p i cellulose acetate (water) or PTFE (ACN) membrane filter before use, and degassed with helium during and before use. Results and Discussion Fig. 1 ( a ) shows a chromatogram corresponding to 2 I pesticides selected for their agricultural interest.The mixture contains organochlorines, triazines, organophosphoru\ compounds. car- barnates and ureic and imidic derivatives with very different polarities. The composition of the mobile phase was optimized by an automated sequential procedure.'"-'x However, over- lyping of peaks occurs if the number of analytes increase\. Fig. l(h) shows a chromatogram containing a new analyte, 19rocyniidone (peak 1 I ) , and overlapping among the peaks of folpet, procymidone and triazophos can be observed. Taking into account the absorption data of the mixture in question, 2 10 nm was first selected as the monitoring wavelength for the detection of the three compounds compromise value. The Table 1 Retenlion time\ of pe4cides i n the multi-residue method Peak ho. Pemctde Re tent i o n ti me/ni i t i 1 2 4 3 0 7 X 9 I 0 I I I' I3 13 1s 16 17 18 I9 2 0 31 22 3 M e t om y I Dimethoate A 1 d i c arb Diclorvos C a r bofu ran Atrnzi tie D i u ron D ich lo ran Mcthiocarb Folpet Procy midone Trinzophos Iproclionc Vinclnzolin Ch lorf'env i nphos Chlorpyrifos methyl Endosulfan sulfate Tetrad ifon I',-Endosul fan oc-Endosulf'm Chlorpyrifos cthyl C'arbophenothion 3.3 3 .1 4.4 5.6 5.2 7.3 8.6 9.9 11.2 13.1 13.4 13.7 13.9 14.7 13.9 16.4 16.7 17.8 18.0 i 8.4 18.7 19.4 0.03 0 0.02 e 2 J r; 0.01 0.00 -11 771 1 T l T-m 5 10 15 T i m h i i n -0.0 I ? Fig. 1 ( a ) C'hromatogram obtained by iri.jection of 20 p1 of pesticide standard solution with a 20 min gradient (2 yg r i l l - - ' of each pesticide at 2 I0 nm). Numbers above the peaks correspond with those given in Table I .( h ) Chromatograni w,ith a new analyte. procymidone (peak number 1 I ) . is observed with 20 min gradient (9 1.18 ml- I of folpet, 4 pg m - k ' of procymidone and 6 png ml I of triazophos).Anolvst, Ortohcv- 1996. Vol. 12 I 1369 R, values are 0.9 for folpet and procymidone and 0.7 for procymidone and triazophos. Table 1 summarizes the retention times of each pesticide. Three-dimensional Spectrochroniatograms The diode-array detector allows the collection of full spectral data at rates of up to several scans per second. With the data it is possible to construct three-dimensional plots of absorbance. wavelength and time. Moreover, these plots can be manipulated to allow the data to be viewed from different angles, including from the end of the chromatogram towards the beginning.The corresponding absorption maxima are located at 226 nm for folpet. at 206 nm for procymidone and at 200 and 245 nm for triazophos (Fig. 2). From the observation of the corresponding absorption \pectra, it i\ evident that folpet and procyniidone present their absorption maxima at close wavelengths, whereas triazophos presents the second maximum absorption at a longer wavelength, but its absorption spectrum overlaps in part with that of procymidone. A potentially more informative way of presenting the chromatograms is to use the cartographic technique of a contour plot, a map of signal intensity in the wavelengtli-time domain (Fig. 3). From this plot it is easier to 4ec the incomplete resolution of folpet, procymidone and triazophos. Because of the highly overlapping peaks, conventional measures of the dillerent analytical signal\ (area or height of chromatographic peaks) cannot be realized. With the aim of resolving the ternary rn i x t u re, a c h e in o m c t r i c a p p ro ac h was e v a I u a t ed .Cross-section Optimization Through the Three-dimensional Data Matrix The contour plots are especially useful i n making cross-sections through the data matrix. in order to pass a s close as possible to the wavelength maxiinurn of each analyte avoiding absorption regions of the others in order to optimize both resolution and sensitivity. Trajectories can be defined, through the contour plot, by the initial and final coordinate (A, t ) pairs. In this work, two trajectories were performed to establish the corresponding cross-sections (Fig.3). In order to select the first linear path, four cross-sections were tested. Their initial coordinates (h, t ) are 200 nm, in the wavelength domain, and in the time domain they are 700,725,750 and 760 s (lines a, b, c and d, respectively, in Fig. 3); the final coordinates (A. t ) are (240. 900) in all cases. 0.60 r\ Wavelengthhm Fig. 2 of folpet and (3) 6 pg nil Absorption spectra of ( I ) 5 pg nil I ofprocymidone, (2) 3 pg inl- I ot triazophos. In the second path the initial and final coordinates (A, t ) are (240, The two-dimensional projections on the wavelength domain, generated by the selected cross-sections through the data matrix, are represented in Fig. 4. The analytical signals obtained 900)-(280, 500).900 800 C 700 d k 600 (I) --- Em i= 400 300 200 100 0 2 -_I-- ---~-- _--.--T . __ - 0 210 220 230 240 250 260 270 : 10 Wavelengthhm Fig. 3 Contour plot ot ( I ) folpcl. ( 2 ) procymidone and ( 3 ) triaiopho\ 'it conceiitiation\ ot 9, 4 and 6 lig nil- 1 , ie\pectively. where the foul trajectories wlected (a, b, c and d) in the f m t linear path optimization of the cro\a-section die plotted. 0.1 2 7- I 1 1 ,- 1 - : . I I T - 2 T 1 I I , ---- r - , ~---i ,__. 1 L - 7---- r T- , ~- 7- 4 200 220 240 260 280 200 220 240 260 280 Waveleng thhm Fig. 4 Two-dimensional projections of the cross-sections produced from the three-dimensional data by plotting absorbance iscr,\us wavelength: ( a ) trajectory a in Fig. 3, ( h ) trajectory b, ( c ) trajectory c and (0) trajectory d.Numbers above the peaks correspond to (1) folpet, ( 2 ) procymidone and (3) triazophos.1370 Analyst, Octohci- 1996, Vol. I2 I after this process are very different from those of the original chromatograms. The four cross-sections tested were selected in order to obtain two-dimensional projections with the best analytical character- istics (resolution and/or sensitivity). It is evident that the four trajectories selected are not the only possibilities. More complicated trajectories, with more than two linear paths, or even non-linear paths. niay be selected for the analysis. In the optimization of the first linear path, triazophos is separated from the other analytes whilst the resolutions between folpet and procymidone are for trajectory a 0.9, b 0.9, c 1.2 and d 1.1.We decided to use the two-dimensional projection obtained from trajectory c as it gave good resolution and sensitivity for the three compounds. In the selection of the second linear path. the initial coordinates (A, t ) are (240, 900) and to select the final coordinates (A, r ) the wavelength values used are 250, 260 and 280 nm, while the time is 500 s in all cases. In Fig. 5 are shown the- trajectories of the selected cross-sections in order to optimize the sensitivity of triazophos. The trajectories tested have little influence in the sensitivity of triazophos (Fig. 6), but in the two-dimensional projection corresponding to Fig. 6(c) the interference due to the peak that appears close to the triazophos peak is avoided. We selected the trajectories defined for the coordinates (A.t ) (200, 750)-(240, 900) for the first path and (240, 900)-(280, SOO) for the second path. In Fig. 7 is presented the isometric projection of the complete spectrochromatogram of the mixture analysed, in which the trajectory of the optimiLed cross-section is marked. In this way, the resolution of the mixture is accomplished, allowing the quantification of each of the arialytes through the adequate calibration lines. 900 800 700 600 ! 5oa i= 4oa 3OC 20c 1 oc c L 10 2 i o 220 230 2 i o 2jo 260 50 : Wavelengthhm 3 Calibration graphs Calibration graphs were obtained from peak heights of two- dimensimal projections for samples of mixture\ of the three compounds, containing different concentrations of folpet, procymidone and triazophos.Good linearity was obtained for all pesticides in the 1.0-10.0 pg ml-I range. Table 2 lists the straight-line equations obtained for the concentration intervals tested and the corresponding statistical parameter values obtained without replicating the experimental points. In order to study the repetitivity of the method, a series of six solutions were prepared, containing 2.0 pg ml- I of folpet, 2.0 pg ml- 1 of procymidone and 2.0 pg ml-l of triazophos, with results of 3.8, 4.5 and 3.l%, respectively, for the RSDs. The values obtained show the high repetitivity of the method. Resolution of Synthetic Ternary Mixtures To validate the method, mixtures of folpet, procymidone and triazophos, in the concentration range I .O-10.0 pg ml- for each pesticide, were prepared, and chromatograms were recorded according to the described procedure.Tablc 3 presents the results of the analysis of different mixtures. Satisfactory 0.08 a, 0.06 0 a 0.04 s: a 0.02 D 0.00 0.08 7- 0 0.00 - -7- - -7- 7 3 220 240 2 3 1 I!), I 0 7 1 + ,.--, - - , - ~ , ~ - - 200 220 240 2 0 280 Wavelengthhm Fig. 6 Two-dimensional projection of the cross-sections produced from the three-diitiensional data by plotting absorbance i'oi-xu.s wavelength: ( ( 1 ) trajectory a i n Fig. 5. (h) trajectory b and ( c ) trajectory c. Numbers above the peaks correspond to ( 1 ) folpel, (2) procymidone and (3) triazophos. Fig. 5 The three tni-jectories selected (a, b and c) in the second linear path optimization of the cross-section, plotted across the contour plot, of the spectrochromatogram of a mixture containing 9 pg nil-' of folpet ( l ) , 4 pg ml- I of procymidone (2) and 6 pg m-1 of triazophos ( 3 ) .Fig. 7 Isometric projection of the spectrochrornatoFrani of the mixture analysed, in which the trajectory of the selected cross-section c is marked.Analyst, October 1996, Vol. 121 1371 Table 2 Calibration graphs for the determination of folpet, procymidone and triazophos by measuring the peak heights at the selected projection of the cross- section Standard Standard error Pesticide Equation* r2 deviation of estimate Folpet y = 1.6154 x 10-3 + 6.6993 X 10V x 0.9973 0.02235 0.009 12 Procymidone y = 1.6105 X lop3+ 1.0103 X lop2 x 0.9973 0.03370 0.01376 Triazophos 4’ = -1.44.52 X 10-’+3.2468 X 10-3 x 0.9993 0.01081 0.00441 * x = Concentration of pesticide in pg ml-1; y = peak height.The calibration graphs were obtained from six experimental points. Table 3 Mean recoveries and RSDs ( n = 5 ) of folpet, procymidone and triazophos in synthetic ternary mixtures Folpet Procymidone Triazophos Recovery RSD Recovery RSD Recovery RSD c* (%) (%) c* (%) (%) c* (%) (% 1 3 91:O 6.2 I 95.8 4.9 4 90.1 6.4 4 105.3 4.4 1 106.3 5.2 7 98.3 5.4 10 106.0 4.8 10 104.9 5.0 10 103.7 5.6 8 96.5 4.5 8 108.4 4.5 8 107.2 3.9 8 90.7 5.0 6 93.5 3.8 7 108.5 5.4 * C = Concentration of pesticide added (pg ml-1). results were obtained, with recoveries ranging from 90.7 to 106.0% for folpet, from 93.5 to 108.4% for procymidone and from 90.1 to 108.5% for triazophos. The results indicate that the complete resolution of the mixture has been accomplished by the proposed approach, showing the high resolving power of the technique.Preconcentration of Pesticides in Water by Solid-phase Extraction (SPE) The proposed method was applied in the determination of pesticides in environmental water samples. A trace enrichment step is necessary to obtain detection limits as low as ppb levels. To evaluate the potential of trace enrichment of the pesticides, samples of ultra-pure water, spiked with 3 pg 1-1 of pesticides, were analysed. The 360 mg Sep-Pak C18 cartridges were conditioned with 5 ml of ACN followed by 5 ml of ultra-pure water without allowing the cartridges to dry out. Water samples of 400 ml were passed through a 0.4 pm filter, connected by PTFE tubes to the conditioned cartridges, at a rate of 8-10 ml min-1; the cartridges were then sucked dry for 5 min. ACN was chosen as solvent for the elution of analytes owing to its suitability for the RP-HPLC system and, finally, 20 p1 were injected.Good linearity was obtained for all substances in the ranges studied (3-11 pg 1-1). The regression coefficients are higher than 0.991 in all cases (M = 7). The detection limits,29 calculated statistically, are 0.3 1,0.29 and 0.43 pg 1- for folpet, proc y midone and tri azophos, respectively . The mean recoveries of the pesticides were 101,98 and 85% for folpet, procymidone and triazophos, respectively. The repeatability in terms of peak height at various concentrations was studied using the conditions described above. The data obtained for 3 pg 1-l indicate that the RSD ranged from 5.5% (triazophos) to 8.4% (procymidone).With groundwaters spiked at a level of 3 pg 1-1, the recoveries were 91, 85 and 83% for folpet, procymidone and triazophos, respectively, and the RSDs were 8.3. 9.1 and 8.7%, respectively. A blank of water without fortification was also analysed in each experiment. The proposed method was applied to the determination of pesticide levels in ground waters from Almeria (Spain) and the chromatograms obtained showed no peaks of the studied pesticides. Conclusions The determination of folpet, procymidone and triazophos mixtures was performed by means of the proposed technique with good repetitivity and sensitivity. The technique is particularly useful for analysing mixtures of analytes in complex samples, as is the case in MR pesticide analysis.The usefulness of the proposed methodology is the resolution of overlapping chromatographic peaks maintaining, at the same time, as much sensitivity in the determination as possible. In addition, the approach would allow a decrease in the time of analysis in certain cases. This can be the case in the separation of several analytes with similar polarities from one with a very different polarity. The method has been applied to the determination of folpet, procymidone and triazophos in water samples at ppb levels with good results. In conclusion, the combination of advanced computational capability with the DAD technology applied in HPLC offers a powerful approach for the resolution of highly overlapping peaks.The authors are grateful to DGICYT (Project PB95-1226) for their financial support. References 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 Edgell, K. W., Erb, E. J., Wesselman, R. J., and Longbottom, J. E., J . AOAC Inf., 1993, 76, 1098. Hernindez, F., Morell, I., Beltran, J., and Lhpez, F. J., Chronzatogra- phia, 1993, 37, 303. Hong, J., Eo, Y., Rhee, J., Kim, T., and Kim. K., J . Chromatogr., 1993,639,261. Edgell, K. W., Erb, E. J., Longbottom, J. E., and L6pez-Avila, V., J . AOAC Int., 1992, 75, 858. de la Colina, C., Baez, M. E., Peiia, A., Romero, E., Dios, G., and Sinchez-Rasero, F., Sci. Total Environ., 1994, 153, 1. Huen, J. M., Gillard, R., Mayer, A. G., Baltensperger, B., and Kern, H., Fresenius’ J . Anal. Chem., 1994, 348, 606. Slobodnik, J., Groenewegen, M.G. M., Brouwer, E. R., Lingeman, H., and Brinkman, U. A. Th., J . Chronzatogr., 1993, 642, 359. Fell, A. F., Clark, B. J., and Scott, H. P., J . Chromatogr., 1984, 316, 423. Gluckman, J. C., Shelly, D. C., and Novotny, M. V., Anal. Chern., 1985, 57, 1546. Wegrzyn, J., Patonay, G., Ford, M., and Warner, I., Anal. Chern., 1990,62, 1754. Garrido Frenich, A., Martinez Galera, M., Gil Garcia, M. D., and Martinez Vidal, J. L., J . Chmmatogr., 1996, 727, 27. Martinez Galera, M., Martinez Vidal, J. L., Garrido Frenich, A., and Gil Garcia, M. D., J . Chronzatogr , 1996, 727, 39. Garrido Frenich, A., Martinez Galera, M., Gil Garcia, M. D., Martinez Vidal, J. L., Mufioz de la Peiia, A., and Salinas, F., J . Chromatogr., submitted for publication. Clark, B. J., Fell, A. F., Scott, H. P., and Westerlund, D. J., J Chromatogr., 1984, 286, 261. Clark, B. J., and Fell, A. F., Cheni. Br., 1987, 23, 1069. Fasanmade, A. A., Fell, A. F., and Scott, H. P., Anal. Chim. Acta, 1986, 187, 233. Fasanmade, A. A., and Fell, A. F., Anal. Chern., 1989, 61, 720. Muiioz de la Peiia, A., Salinas, F., Galeano, T., and Guiberteau, A., Anal. Chim. Acta, 1990, 234, 263. Parrilla, P., Martinez Galera, M., Martinez Vidal, J. L., and Garrido Frenich, A., Analyst, 1994, 119, 2231.1372 Analyst, October 1996, Vol. I21 20 21 22 23 24 25 Dimuccio, A., Girolimetti, S., Ausili, A., Ventriglia, M., Generali, T., and Vergori, L., J . Chromatogr., 1993, 643, 363. Holland. T. P., Naughton, E. D., and Malcolm, P. C., J. AOAC Int., 1994, 77, 79. Ogawa, M., Ohtsubo, T., Tsuda, S., and Tsuji, K., J . AOAC Int., 1993, 76, 83. Liao, W., Joe, T., and Cusick, G., J . AOAC Int., 1991, 74, 554. Parrilla, P., Martinez Vidal, J. L., and Fernandez Alba, A. R., J . Liq. Chromutogr., 1993, 16, 4019. Parrilla, P., Martinez Vidal, J. L., Martinez Galera, M., and Frenich, A. G., Fwseniub’ J . Anal. Chern., 1994, 350, 633. 26 27 28 29 Bellar, T. A., and Budde, W. L., Anal. Chem., 1988, 60, 2076. GRAPHER and SURFER for Windows Software Package Version 5.0, Golden Software, CO, 1994. Martinez Vidal, J. L., Parrilla, P., Fernandez Alba, A. R., Carreiio, R., and Herrera, F., J . Liq. Chromatogr., 1995, 18, 2969. Long, G. L., and Winefordner, J. D., Anal. Chem., 1983, 55, 713. Paper 6102345B Received April 3, I996 Accepted June 7, 1996